基于图像特征角点的无人机停机坪识别算法

Recognition Algorithm of the Apron for Unmanned Aerial Vehicle Based on Image Corner Points

为了实现无人机自主导航控制,设计了一种面向四旋翼飞行器对停机坪地标多角度观测的识别算法。首先,将停机坪地标图像分为特征圆及特征字符H两部分,用Hough算法识别特征圆。接着,在特征字符H识别过程中,发现归一化后的H字符角点间存在确定的几何关系,且在一定范围内,该几何关系对飞行器位姿及周围环境敏感度较低。因此,将特征角点识别算法应用于停机坪地标图像的特征字符识别,整个算法在不增加额外运算开销的情况下提高了识别效率。实验结果表明,基于图像特征角点的无人飞行器停机坪地标识别算法,在识别准确性和实时性方面都要优于基于线性特征的识别算法。同时满足了飞行器自主导航系统稳定可靠、精度高、抗干扰能力强等要求。

To solve the problem of autonomous navigation control of the quad rotor, a new recognition algorithm of landmark with multi angle observations is proposed. According to the characteristics of landmark, the image is divided into two parts with circle and character H. The analysis reveals a geometric correlation exits between comer points of character H. The correlation has a low sensitivity to rotation post and environment. The experiments show that recogni- tion algorithm based on the image comer points not only keeps the real time performance but also detects landmarks at any degree of rotation in the 2D image plane from the air, by comparing it with linear feature algorithms. The results of recognition experiments verify the effectiveness of the approach, and the algorithms meets the demands of autonomous navigation control of unmanned aerial vehicle （UAV） based on the quadrotor with such advantages as high accuracy, good reliability and resistance to interference.